Page 8 of 25
Re: CGI tech news box
Posted: Thu Nov 15, 2018 4:27 pm
by kintuX
Psychopath Renderer: A different Approach
I've been away from Psychopath for a while working on other projects, but recently I stumbled upon a blog post by Yining Karl Li, entitled "Mipmapping with Bidirectional Techniques". In it, he describes his solution to a problem I've been pondering for a while: how to handle texture filtering in the context of bidirectional light transport...
--------------------------------------------------------------
Oh, and especially good news for
lacilaci
Optix Denoiser with Cycles... soon.
Re: CGI tech news box
Posted: Wed Nov 28, 2018 10:43 pm
by kintuX
A radiative transfer framework for non-exponential media
Benedikt Bitterli, Srinath Ravichandran, Thomas Müller, Magnus Wrenninge, Jan Novák, Steve Marschner, Wojciech Jarosz. A radiative transfer framework for non-exponential media. ACM Transactions on Graphics (Proceedings of SIGGRAPH Asia), 37(6):225:1–225:17, November 2018.
Even in photography, such a clear image of rain falling from the cloud with its backgrounds is quite hard and expensive to make.
Re: CGI tech news box
Posted: Sun Dec 02, 2018 9:54 pm
by Sharlybg
Unity global instant GI without precomputation :
https://forum.unity.com/threads/segi-fu ... on.410310/
SEGI Opensource Project :
Re: CGI tech news box
Posted: Mon Dec 03, 2018 9:46 am
by Sharlybg
Some Good boost to Bidir engine with such kind of boost Bidir could become a standar for daily rendering workflow :
https://www.researchgate.net/publicatio ... _Rendering
Re: CGI tech news box
Posted: Mon Dec 03, 2018 10:26 am
by Dade
I tried this in old Lux years ago:
heterogeneous hybrid CPU/GPU rendering has too much problems. Note the
heterogeneous, it means you are assigning to CPU and GPU different type tasks and you get the end result when all is done. This is different from our current
homogeneous hybrid rendering where we assign to CPU and GPU the same type of task.
Short version: you have an infinitely fast GPU (it does everything in 0 secs) and a normal CPU. We have 2 task types A and B. Heterogeneous hybrid rendering will assign A tasks to the GPU and B tasks to the CPU. The time to do everything will be cpu time for B ... even with an infinitely fast GPU.
Homogeneous hybrid rendering will render everything in 0 secs thanks to the infinitely fast GPU.
Re: CGI tech news box
Posted: Sun Jan 20, 2019 11:48 am
by epilectrolytics
Re: CGI tech news box
Posted: Wed Jan 23, 2019 5:17 pm
by Sharlybg
Re: CGI tech news box
Posted: Wed Jan 23, 2019 5:57 pm
by lacilaci
interesting. So they have the same problem with photon mapping as other renderers (splotches, too many settings and gpu memory, but working caustics).
I wonder if Dade's solution will be first one that overcomes the usual issues. I mean last post of his experiments were splotch free and memory usage wasn't terrible...
Re: CGI tech news box
Posted: Wed Jan 23, 2019 7:59 pm
by Dade
lacilaci wrote: ↑Wed Jan 23, 2019 5:57 pm
interesting. So they have the same problem with photon mapping as other renderers (splotches, too many settings and gpu memory, but working caustics).
I wonder if Dade's solution will be first one that overcomes the usual issues. I mean last post of his experiments were splotch free and memory usage wasn't terrible...
I don't feel the pressure
Re: CGI tech news box
Posted: Thu Jan 24, 2019 12:46 pm
by Mango3
interesting. So they have the same problem with photon mapping as other renderers (splotches, too many settings and gpu memory, but working caustics).
Standard photon mapping is a robust and well known method but has been largely superseded by far better methods (progressive photon mapping -> stochastic progressive photon mapping) because it cannot handle distributed ray tracing effects like motion blur, depth-of-field or anti-aliasing.
If you then couple bidirectional path-tracing with photon mapping for what both are best then you end up with vertex connection merging (vcm).
RenderMan has it for instance.